The LOOC results corresponded well with the
above hypothesis: learning based on similar characteristic
fac-tors as the validating data results in better
accuracy. In other word, the flood predicting model
cannot be applied universally, unless both the training
and testing attributes share the similar characteristics